[comp.parallel] Pseudospectral codes on the CM-II

jek@cs.duke.edu (James E. Kittock) (05/17/91)

I am working on a project to port a 2d mixed finite-difference
spectral code from the CRAY YMP to a CM-II. The bottlenecks of
the code involve two-dimensional FFT's (nx one-dimensional FFT's
on vectors of length ny), in which products of fields are
evaluated by the pseudospectral methods, and solution of pentadiagonal
banded matrices (nx matrices of size ny x ny where each square
matrix is band 5).

I would greatly appreciate any code or references or both that
would help me in this port. If there is interest, I will summarize
to the net.

	Thanks,

	James Kittock

P.S. nx and ny are expected to be of order 256 to 1024 in size.

-- 
=========================== MODERATOR ==============================
Steve Stevenson                            {steve,fpst}@hubcap.clemson.edu
Department of Computer Science,            comp.parallel
Clemson University, Clemson, SC 29634-1906 (803)656-5880.mabell

jos@Think.COM (05/20/91)

Hi James,

I have some knowledge about pseudospectral codes because of
the work done in that area in my group at Stanford. The person
who did the work is John Etgen (zjte01@cmns.think.com).

I'm now working at Thinking Machines, and I have implemented
a spectral algorithm that involves the 2-D FFTs you're mentioning.
I've also worked on a finite-difference code that solves a tridiagonal
system (we have a fast tridiagonal solver in our scientific library).

I'd be happy to help you with any questions you may have.

Jos
-------

Article 2604 (1 more) in comp.parallel (moderated):
From: jek@cs.duke.edu (James E. Kittock)
Subject: Pseudospectral codes on the CM-II...
Date: 16 May 91 18:50:50 GMT



I am working on a project to port a 2d mixed finite-difference
spectral code from the CRAY YMP to a CM-II. The bottlenecks of
the code involve two-dimensional FFT's (nx one-dimensional FFT's
on vectors of length ny), in which products of fields are
evaluated by the pseudospectral methods, and solution of pentadiagonal
banded matrices (nx matrices of size ny x ny where each square
matrix is band 5).

I would greatly appreciate any code or references or both that
would help me in this port. If there is interest, I will summarize
to the net.

        Thanks,

        James Kittock

P.S. nx and ny are expected to be of order 256 to 1024 in size.

--